From: Jonathan Campbell <jonathan@castus.tv>
Date: Sun, 17 Apr 2016 05:29:15 +0000 (-0700)
Subject: let's show Sparky4 what I mean by vertical column font rendering.
X-Git-Url: http://4ch.mooo.com/gitweb/?p=16.git;a=commitdiff_plain;h=c1d3221afb5fe541d8c886ff66198329c153e48e

let's show Sparky4 what I mean by vertical column font rendering.
this code should demonstrate what I mean about minimizing our writes to
the map mask register and using arithmetic to step from byte to byte.
at least in DOSBox-X, the characters appear much faster now.
---

diff --git a/pcx2vrl b/pcx2vrl
index 3d9ac6e6..06179c75 100755
Binary files a/pcx2vrl and b/pcx2vrl differ
diff --git a/pcxsscut b/pcxsscut
index d743ac3b..2ce56d28 100755
Binary files a/pcxsscut and b/pcxsscut differ
diff --git a/src/lib/modex16.c b/src/lib/modex16.c
index 4e7004e1..bada2e67 100755
--- a/src/lib/modex16.c
+++ b/src/lib/modex16.c
@@ -914,7 +914,8 @@ void modexprint(page_t *page, word x, word y, word t, word col, word bgcol, cons
 		JNZ L1
 	}
 //TODO: OPTIMIZE THIS!!!!
-		modexDrawCharPBuf(page, x, y, t, col, bgcol, addrq);
+		modexDrawCharPBuf(page, x/*for mode X planar use*/, y/*redunant, remove*/, t, col, bgcol, addrq);
+		addrq += 2; /* move 8 pixels over (2 x 4 planar pixels per byte) */
 
 		//if(!q) getch();
 	}
diff --git a/src/lib/modex16.h b/src/lib/modex16.h
index c1c45252..f85eb055 100755
--- a/src/lib/modex16.h
+++ b/src/lib/modex16.h
@@ -71,7 +71,7 @@ static struct pcxHeader {
 
 /* -========================== Types & Macros ==========================- */
 #define PAGE_OFFSET(x,y) (((y)<<6)+((y)<<4)+((x)>>2))
-#define PLANE(x) (1<< (x&3))
+#define PLANE(x) (1 << ((x) & 3))
 #define SELECT_ALL_PLANES() outpw(0x03c4, 0xff02)
 #define     PALSIZE            768
 
@@ -149,6 +149,8 @@ void modexPalOverscan(byte *p, word col);
 void modexchkcolor(bitmap_t *bmp, word *q, word *a, word *aa, word *z, word *i/*, word *offset*/);
 void modexputPixel(page_t *page, int x, int y, byte color);
 byte modexgetPixel(page_t *page, int x, int y);
+
+#if 0 // not needed anymore. maybe good for reference purposes though.
 static inline void modexwritepixel(page_t *page, int x, int y, word addr, byte color)
 {
 	/* Each address accesses four neighboring pixels, so set
@@ -172,6 +174,8 @@ static inline byte modexreadPixel(page_t *page, int x, int y, word addr)
 	outpw(GC_INDEX+1, x & 3);
 	return vga_state.vga_graphics_ram[addr];
 }
+#endif
+
 void modexprint(page_t *page, word x, word y, word t, word col, word bgcol, const byte *str);
 void modexprintbig(page_t *page, word x, word y, word t, word col, word bgcol, const byte *str);
 void modexpdump(page_t *pee);
diff --git a/src/lib/modex16/16render.c b/src/lib/modex16/16render.c
index fe78022d..8e6fa436 100755
--- a/src/lib/modex16/16render.c
+++ b/src/lib/modex16/16render.c
@@ -435,34 +435,38 @@ modexDrawSpritePBufRegion(page_t *page, int x, int y,
     }
 }
 
-void modexDrawCharPBuf(page_t *page, int x, int y, word t, word col, word bgcol, word addr)
+void modexDrawCharPBuf(page_t *page, int x/*for planar selection only*/, int y/*redundant, remove*/, word t, word col, word bgcol, word addr)
 {
-	word i, j, k;
-	for(i=0; i<romFonts[t].charSize; i++)
-	{
-		j=1<<8;
-		k=0;
-		//every "pixel" row
-		while(j)
-		{
-// 			if(q)
-			//_fmemcpy(page->data + (((page->width/4) * (y+i)) + ((x+romFontsData.chw+k) / 4)), romFontsData.l[i] & j ? col:bgcol, 2);
-			modexwritepixel(page, x+k+romFontsData.chw, y+i, addr, romFontsData.l[i] & j ? col:bgcol);
-// 			else
-				//printf("l[i]=%c j=%02u l[i] & j=%02u %c\n", l[i] , j, l[i] & j, l[i] & j ? '*':' ');
-				//printf("%c", l[i] & j ? '*':' ');
-// 				romFontsData.z[k]=romFontsData.l[i] & j ? '*':' ';
-			j>>=1;
-			k++;
+	/* vertical drawing routine by joncampbell123.
+	 *
+	 * optimize for VGA mode X planar memory to minimize the number of times we do I/O write to map mask register.
+	 * so, we enumerate over columns (not rows!) to draw every 4th pixel. bit masks are used because of the font bitmap.
+	 * 
+	 * NTS: addr defines what VGA memory address we use, "x" is redundant except to specify which of the 4 pixels we select in the map mask register. */
+	word drawaddr;
+	word colm, row;
+	byte fontbyte;
+	byte plane;
+	byte m1,m2;
+
+	plane = x & 3;
+	m1 = 0x80; // left half
+	m2 = 0x08; // right half
+	for (colm=0;colm < 4;colm++) {
+		drawaddr = addr;
+		modexSelectPlane(PLANE(plane));
+		for (row=0;row < 8;row++) {
+			fontbyte = romFontsData.l[row];
+			vga_state.vga_graphics_ram[drawaddr  ] = (fontbyte & m1) ? col : bgcol;
+			vga_state.vga_graphics_ram[drawaddr+1] = (fontbyte & m2) ? col : bgcol;
+			drawaddr += page->width >> 2;
+		}
+
+		m1 >>= 1;
+		m2 >>= 1;
+		if ((++plane) == 4) {
+			addr++;
+			plane = 0;
 		}
-// 		if(!q)
-// 		{
-// 			for(k=0;k<9;k++)
-// 			{
-// 				printf("%c", romFontsData.z[k]);
-// 			}
-// 			printf("\n");
-// 		}
 	}
-	romFontsData.chw += k;
 }
diff --git a/vrl2vrs b/vrl2vrs
index acdd4bda..cee2d1d5 100755
Binary files a/vrl2vrs and b/vrl2vrs differ
diff --git a/vrsdump b/vrsdump
index 8ba0497d..a75a903a 100755
Binary files a/vrsdump and b/vrsdump differ